DESeq Analysis: PEA2 vs PEA1

cont <- params$controlCellLine
exp <- params$experimentalCellLine
countmatrix.all <- params$countmatrix.all
metadata.all <- params$metadata.all
rm(params) # Remove the parameters so that we can make subsequent parameterized calls

Perform analysis of PEA2 vs PEA1

Gather and organize raw data

metadata.pair <- as.data.frame(metadata.all) %>%
  filter(CellLine == cont | CellLine == exp)
as.data.frame(metadata.pair)

countmatrix.pair <- countmatrix.all[, metadata.pair$ShortName]
as.data.frame(countmatrix.pair)

Differential expression analysis

Run Deseq on the data set.

# Saving time by just loading the dds we already ran (recent changes are all after this point)
load(str_interp("Rdata/${exp}_vs_${cont}_dds.RData"))

Print DESeq2 results

res <- results(dds.pair, contrast = c("CellLine", exp, cont), alpha = 0.05)
res <- res[order(res$log2FoldChange), ]
outFile <- str_interp("output/${exp}_vs_${cont}_deseq_results.csv")
write.csv(as.data.frame(res), file = outFile)

Filter DESeq2 results for significant genes

Filter res for padj < 0.05

res.filtered <- as.data.frame(res) %>%
  filter(padj < 0.05)
  # filter(log2FoldChange >= 1.5 | log2FoldChange <= -1.5)
res.filtered <- res.filtered[order(res.filtered$log2FoldChange, decreasing = TRUE),]
res.filtered

Significant “Up” genes (PEA2 compared to control PEA1)

up.unfiltered <- subset(res, log2FoldChange > 0)
up.unfiltered <- up.unfiltered[order(up.unfiltered$log2FoldChange, decreasing = TRUE), ]
outFile <- str_interp("output/${exp}_vs_${cont}_all_upregulated_genes.csv")
write.csv(up.unfiltered[, c("log2FoldChange", "padj")], file = outFile)
up.unfiltered[, c("log2FoldChange", "padj")]

log2 fold change (MLE): CellLine PEA2 vs PEA1 
 
DataFrame with 9935 rows and 2 columns
        log2FoldChange        padj
             <numeric>   <numeric>
SULT1A3        23.0809 2.51448e-08
EHF            16.7101 3.10531e-42
CCDC129        14.9666 6.21388e-34
AGR2           14.4740 1.10314e-31
ESRP1          14.2988 5.77523e-31
...                ...         ...
FLYWCH1    8.04855e-04    0.998286
RN7SK      5.72356e-04    0.999080
ISG20L2    5.70131e-04    0.998178
CYB5B      5.51788e-04    0.998286
MYO5A      7.72654e-05    0.999854

up <- subset(res.filtered, log2FoldChange > 0)
up <- up[order(up$log2FoldChange, decreasing = TRUE), ]
outFile <- str_interp("output/${exp}_vs_${cont}_significantly_upregulated_genes.csv")
write.csv(up[, c("log2FoldChange", "padj")], file = outFile)
print(up[, c("log2FoldChange", "padj")])

Significant “Down” genes (PEA2 compared to control PEA1)

down.unfiltered <- subset(res, log2FoldChange < 0)
down.unfiltered <- down.unfiltered[order(down.unfiltered$log2FoldChange, decreasing = FALSE), ]
outFile <- str_interp("output/${exp}_vs_${cont}_all_downregulated_genes.csv")
write.csv(down.unfiltered[, c("log2FoldChange", "padj")], file = outFile)
print(down.unfiltered[, c("log2FoldChange", "padj")])

log2 fold change (MLE): CellLine PEA2 vs PEA1 
 
DataFrame with 11299 rows and 2 columns
             log2FoldChange        padj
                  <numeric>   <numeric>
IPO11-LRRC70       -23.0428 2.66766e-08
LOC101929304       -22.2940 7.88754e-08
TM4SF18            -14.9989 3.51917e-34
FAT1               -14.6631 4.91444e-20
TRNAD17            -12.9771 2.87443e-03
...                     ...         ...
COX8A          -1.00084e-03    0.996937
NELFE          -5.88644e-04    0.998178
WASH5P         -5.15565e-04    0.998848
RPN2           -3.55934e-04    0.998848
SDHAF2         -1.90322e-05    0.999942

down <- subset(res.filtered, log2FoldChange < 0)
down <- down[order(down$log2FoldChange, decreasing = TRUE), ]
outFile <- str_interp("output/${exp}_vs_${cont}_significantly_downregulated_genes.csv")
write.csv(down[, c("log2FoldChange", "padj")], file = outFile)
print(down[, c("log2FoldChange", "padj")])

Volcano Plot

as.data.frame(res) %>%
  ggplot(aes(x = log2FoldChange, y = -log10(padj), label = rownames(res))) +
  geom_point() +
  theme_minimal() +
  scale_color_manual(values = c("black", "blue", "red")) +
  geom_text_repel() +
  geom_hline(yintercept = 1.301) +
  geom_vline(xintercept = 1.2) +
  geom_vline(xintercept = -1.2) +
  xlim(-10, 10)

Warning: Removed 11833 rows containing missing values (`geom_point()`).

Warning: Removed 11833 rows containing missing values (`geom_text_repel()`).

Warning: ggrepel: 17884 unlabeled data points (too many overlaps). Consider increasing
max.overlaps

Gene Ontology

GSEA of genes differentially express in (PEA2 compared to control PEA1)

Perform gene set enrichment analysis using Cluster Profiler. This gives us GO pathways that are significantly regulated based on the log2fold change of expression of individual genes.

Using a pvalue Cutoff of 0.05

gene_list <- res$log2FoldChange
names(gene_list) <- rownames(res)
gene_list <- sort(gene_list, decreasing = TRUE)

# Set the seed so our results are reproducible:
set.seed(2023)
gsea_res <- gseGO(gene_list, ont = "BP", OrgDb = "org.Hs.eg.db", keyType = "SYMBOL", seed = TRUE, pvalueCutoff = 0.05)

preparing geneSet collections...

GSEA analysis...

Warning in preparePathwaysAndStats(pathways, stats, minSize, maxSize, gseaParam, : There are ties in the preranked stats (13.32% of the list).
The order of those tied genes will be arbitrary, which may produce unexpected results.

Warning in fgseaMultilevel(pathways = pathways, stats = stats, minSize = minSize, : For some
pathways, in reality P-values are less than 1e-10. You can set the `eps` argument to zero for
better estimation.

leading edge analysis...

done...

# Format output
gsea_res_df <- as.data.frame(gsea_res)
gsea_res_df <- gsea_res_df %>%
  mutate(original_row_num = row_number())
gsea_res_df <- gsea_res_df[order(gsea_res_df$NES, decreasing = TRUE),]
row.names(gsea_res_df) <- gsea_res_df$ID

NES is the normalized enrichment score.

gsea_res_df_short <- gsea_res_df[c("pvalue", "p.adjust", "NES", "Description")]
gsea_res_df_short$"core_enrichment_genes" <- gsea_res_df$core_enrichment

Upregulated pathways

gsea_res_df_short.up <- subset(gsea_res_df_short, gsea_res_df_short$NES >= 0)
outFile <- str_interp("output/${exp}_vs_${cont}_significantly_upregulated_pathways.csv")
write.csv(gsea_res_df_short.up, file = outFile)
gsea_res_df_short.up

GSEA plot of the five most upregulated pathways (or least downregulated)

maxIndex <- min(5, nrow(gsea_res_df)) # Prevents us from trying to access out of bounds if there are not five pathways
top5PathwaysIds = gsea_res_df[1:maxIndex, "original_row_num"]

gseaplot2(gsea_res, geneSetID = top5PathwaysIds, pvalue_table = FALSE, ES_geom = "dot")

Volcano Plot (Average NES & adjusted p value)

as.data.frame(gsea_res_df_short.up) %>%
  ggplot(aes(x = NES, y = -log10(p.adjust), label = rownames(gsea_res_df_short.up))) +
  geom_point() +
  theme_minimal() +
  scale_color_manual(values = c("black", "blue", "red")) +
  geom_text_repel() +
  geom_hline(yintercept = 1.301) +
  geom_vline(xintercept = 1.2) +
  geom_vline(xintercept = -1.2) +
  xlim(-10, 10)

Warning: ggrepel: 80 unlabeled data points (too many overlaps). Consider increasing
max.overlaps

Downregulated pathways

gsea_res_df_short.down <- subset(gsea_res_df_short, gsea_res_df_short$NES <= 0)
outFile <- str_interp("output/${exp}_vs_${cont}_significantly_downregulated_pathways.csv")
write.csv(gsea_res_df_short.down, file = outFile)
gsea_res_df_short.down

GSEA plot of the five most downregulated pathways (or least upregulated)

minIndex <- max(1, nrow(gsea_res_df) - 5) # Prevents us from trying to access out of bounds if there are not five downregulated pathways
bottom5PathwaysIds = gsea_res_df[minIndex:nrow(gsea_res_df), "original_row_num"]
gseaplot2(gsea_res, geneSetID = bottom5PathwaysIds, pvalue_table = FALSE, ES_geom = "dot")

Volcano plot (Average NES & adjusted p value)

as.data.frame(gsea_res_df_short.down) %>%
  ggplot(aes(x = NES, y = -log10(p.adjust), label = rownames(gsea_res_df_short.down))) +
  geom_point() +
  theme_minimal() +
  scale_color_manual(values = c("black", "blue", "red")) +
  geom_text_repel() +
  geom_hline(yintercept = 1.301) +
  geom_vline(xintercept = 1.2) +
  geom_vline(xintercept = -1.2) +
  xlim(-10, 10)

Warning: ggrepel: 10 unlabeled data points (too many overlaps). Consider increasing
max.overlaps

Clustered pathways

Clustered upregulated pathways

Use Revigo to cluster upregulated pathways

revigo_input.cellline.up <- gsea_res_df_short.up[c("p.adjust")]
rownames(revigo_input.cellline.up) <- rownames(gsea_res_df_short.up)

simMatrix <- calculateSimMatrix(rownames(revigo_input.cellline.up),
  orgdb = "org.Hs.eg.db",
  ont = "BP",
  method = "Rel"
)

preparing gene to GO mapping data...

preparing IC data...

scores <- setNames(-log10(revigo_input.cellline.up$p.adjust), rownames(revigo_input.cellline.up))

if (nrow(revigo_input.cellline.up) > 1) {
  reducedTerms <- reduceSimMatrix(simMatrix,
    scores,
    threshold = 0.7,
    orgdb = "org.Hs.eg.db"
  )
} else {
  reducedTerms <- data.frame(matrix(ncol = 0, nrow = 0))
  print("There will be no graphs appearing below this because there were not enough significantly upregulated pathways to meaningfully cluster them")
}

Revigo interactive scatter plot. Distances represent the similarity between terms, axes are the first 2 components of a PCA plot, Each bubble indicates the representative (chosen mostly by p-value) from a cluster of terms. Size of the bubble indicates the generality of the term (large meaning a more general term).

if (nrow(reducedTerms) > 2) {
  revigo_scatterplot(simMatrix, reducedTerms)
}

Revigo heatmap plot. Similar terms clustered

if (nrow(reducedTerms) > 2) {
  heatmapPlot(simMatrix,
    reducedTerms,
    annotateParent = TRUE,
    annotationLabel = "parentTerm",
    fontsize = 6
  )
}

This is the same content, but interactive.

if (nrow(reducedTerms) > 2) {
  revigo_heatmap(simMatrix, reducedTerms)
}

Warning: Specifying width/height in layout() is now deprecated.
Please specify in ggplotly() or plot_ly()

Revigo treemap plot. Terms grouped/colored based on parent. Space is proportional to statistical significance of the GO term (-log10(pvalue)).

if (nrow(reducedTerms) > 2) {
  treemapPlot(reducedTerms)
}

Clustered downregulated pathways

Use Revigo to cluster downregulated pathways

revigo_input.cellline.down <- gsea_res_df_short.down[c("p.adjust")]
rownames(revigo_input.cellline.down) <- rownames(gsea_res_df_short.down)

simMatrix <- calculateSimMatrix(rownames(revigo_input.cellline.down),
  orgdb = "org.Hs.eg.db",
  ont = "BP",
  method = "Rel"
)

preparing gene to GO mapping data...

preparing IC data...

scores <- setNames(-log10(revigo_input.cellline.down$p.adjust), rownames(revigo_input.cellline.down))

if (nrow(revigo_input.cellline.down) > 1) {
  reducedTerms <- reduceSimMatrix(simMatrix,
    scores,
    threshold = 0.7,
    orgdb = "org.Hs.eg.db"
  )
} else {
  reducedTerms <- data.frame(matrix(ncol = 0, nrow = 0))
  print("There will be no graphs appearing below this because there were not enough significantly downregulated pathways to meaningfully cluster them")
}

Revigo interactive scatter plot. Distances represent the similarity between terms, axes are the first 2 components of a PCA plot, Each bubble indicates the representative (chosen mostly by p-value) from a cluster of terms. Size of the bubble indicates the generality of the term (large meaning a more general term).

if (nrow(reducedTerms) > 2) {
  revigo_scatterplot(simMatrix, reducedTerms)
}

Revigo heatmap plot. Similar terms clustered

if (nrow(reducedTerms) > 2) {
  heatmapPlot(simMatrix,
    reducedTerms,
    annotateParent = TRUE,
    annotationLabel = "parentTerm",
    fontsize = 6
  )
}

This is the same content, but interactive.

if (nrow(reducedTerms) > 2) {
  revigo_heatmap(simMatrix, reducedTerms)
}

Warning: Specifying width/height in layout() is now deprecated.
Please specify in ggplotly() or plot_ly()

Revigo treemap plot. Terms grouped/colored based on parent. Space is proportional to statistical significance of the GO term (-log10(pvalue)).

if (nrow(reducedTerms) > 2) {
  treemapPlot(reducedTerms)
}